FeelTech FY6600 60MHz 2-Ch VCO Function Arbitrary Waveform Signal Generator

[soundtec]

 I switched the output frequency to zero on my unit and scaled through the different amplitude  ranges 0-.6, .6-5volts,and 5-20 volts ,I measured the output noise on the high range,5.1volts or above (3095 engaged)to be 150microvolts ac rms (0.00015volts),that seems to translate into around 100db dynamic range refered to 20 volts output .I couldnt measure directly the noise on the two lower ranges with the meter . It would be interesting to compare this figure to other user's results .

I also monitored the output through a mixer with headphones ,there was no trace of hum or any kind of digital artifact
in the noise floor ,just the usual random resistance noise ,a good bit above the mixer noise floor. Be carefull not to touch the frequency setting while listening to the backround noise in the head phones *
I might give the original smps one last try for a comparative test for noise level at the outputs.
For now with the lid on and both channels 20 volts output into 50 ohms ,the transformer is starting to heat up a bit .

[soundtec]

Im quite surprised by the amount of heat Im getting from the transformer ,its not burning up or anything ,but its certainly working towards the upper end of its capacity when the unit is driven . The fact that the enclosure is made of plastic means the transformer cant loose heat to it in the way it does with a metal box . I was thinking of adding an extra aluminium plate on the under side of the case ,secured in by the four bolts on the transformer ,just to try and loose a little more heat to the outside world ,I think the lower corner vents could breathe better as well  so Ill open these out a bit  .The dividing screen I put in effectively seperates the two sections thermally but doesnt help reduce the temp in the psu section much. A lot  of the extra capacity of the new psu ends up  having to be dumped as heat in the transformer and regulators ,but thats the cost of quiet supply rails I guess.

I think I read way back in the thread somewhere that 5 volts or below is a bit marginal ,if we make 3.3v from the five that only leaves 1.7 volts drop on the regulator ,maybe 5.3 volts might be better here .Wont make a huge difference to the temp of the 5volt supply board in any case ,I might need to find a way to thermally bond it to the metal plate that supports it ,while maintaining ground isolation  . The tabs on the regulator chips have their own isolated heat sinking sections on the back of the board ,but it seems these arent adequate in this instance , I need to find a way to mount the board and thermally couple the bridge and the ic's to the plate ,Any ideas welcome.
I'll post a couple of properly lit photos shortly  to illustrate better .

[bugi]

Im quite surprised by the amount of heat Im getting from the transformer ,its not burning up or anything ,but its certainly working towards the upper end of its capacity when the unit is driven.

Have you checked the math (or measured) what is the conduction angle on the transformer?  That is, not too big filter caps so that the transformer can work during longer periods of the cycle leading to lower losses in it.  It is a compromise, though..

I think I read way back in the thread somewhere that 5 volts or below is a bit marginal ,if we make 3.3v from the five that only leaves 1.7 volts drop on the regulator ,maybe 5.3 volts might be better here.

This depends a lot on the particular regulator and the operating (max) current. For example, I have (currently only) plans where a linear regulator should end up working at around 80mV drop, when its basic advertised-single-parameter drop is 0.5V. (Yeah, low drop to start with, but I'm going to drive it at less than 10% of maximum current, and the datasheet was nice enough to characterize lower currents, too).  Then again, these kind of devices tend to have tiny regulator chips, and design leaves them to run near the max currents.

If you're already using DC-DC for the 5V generation, there isn't much more to gain in that. I'd concentrate on optimizing the higher voltage rails, and since without DC-DC the total regulation drop there decides the total heat to be wasted, and it is more about choosing how to spread it among components... Except for the transformer conduction angle; if you can allow more ripple before regulator (and possibly lower output voltage), that might help with the transformer heating and improve total efficiency.  Judging from the picture, there is only one (large) cap per rail, so the filter capacitance isn't likely too huge, but worth checking, perhaps.

[DaveR]

Hi soundtec,

I think the heat from your transformer is probably inherent in its construction (E-I type, lots of iron, lots of eddy currents, lots of waste heat).  For a comparison, I've just run mine for 30 minutes with both channels at 9MHz, 20Vpp into 50 ohms and the heat produced by the PS is:

R-core transformer (30VA, 15-0-15v, 9v, 0.5kg) - nothing that can be felt after removing the lid.
Regulators on 12-15v board (same type as yours) - barely warm (running at 13.5v).
Regulator on 5v board - hot (normally runs about 70C).
Case - still cold above the transformer.

This probably says a lot for the efficiency of R-cores over E-I types!  This project is the first time I've ever used one (or even heard of them, actually - I only came across them while searching for a suitable "regular" type), but they'll certainly be my first choice for similar jobs in the future.

Just some food for thought.

Dave

[plb]

Im quite surprised by the amount of heat Im getting from the transformer

I'd guess it's a cheaper transformer with a marginal amount of iron for the primary voltage. The core is probably saturating at peak, causing a current spike, hence dissipation, in the primary winding.

[soundtec]

Thanks Bugi ,Dave and PLB.
I kept the smoothing caps a reasonable size ,2x3300uf @ 50volts on the +/- rails ,and just 1000uf on the 5 volt rail . Electrically Im happy with it ,I am scrubbing off some extra volts on the regulators ,so I expect them to run  a bit hot. I think the reason the transformer heats a lot is theres no where for it to loose heat too, such as a metal chassis and open air to conduct the heat away . Interesting point you made Bugi about it being a matter of sharing the heat around among the different components , I could try some different low voltage drop diodes /bridges ,loose some dissipation in the transformer ,but then I end up having to push the regulators even harder due to the extra volts.

It takes an hour  before things start getting warm on my unit ,but it continues to rise upto four or five hours before leveling out, even with both channels in standby . Just shifting a little heat outside the box might be enough to keep things under better control ,I'll try the metal radiator plate on the underside of the case and see if that helps.

[nctnico]

Im quite surprised by the amount of heat Im getting from the transformer ,its not burning up or anything ,but its certainly working towards the upper end of its capacity when the unit is driven .

That is a typical problem with low power transformers: many get very hot. In one of my designs I replaced a transformer with a switching one which fits in the same footprint.

[Mr. Scram]

What happened to I think it was rhb's ultimatum? Has FeelTech responded?

[SMB784]

Soldering the TXCO modul is something you need patience for. I placed kapton tape on the bottom side of the module so that all the not needed pads were covered.

On the TXCO module, what do you do about pin 8? Do you cover it with kapton or solder it together with Vcc?

[Candid]

I only connected pin 4, 5 and 9. All others are covered by kapton tape.

[SMB784]

I only connected pin 4, 5 and 9. All others are covered by kapton tape.

That is interesting, I wonder how it affects the operation of the device to leave the trip state buffer enable/disable pin floating. It doesn't mention that anywhere in the datasheet

[soundtec]

Well it looks like an Fy6800 is on its way out to me .
I didnt actually ask if I get to keep the unit ,but I think that is the intention at the end of the day .
Of course I still very much hope the handfull with broken units get satifsfaction on repair or replacement ,and I did mention this in my reply to Feeltech also.
It'll be interesting to compare the units side by side and see exactly what changes have been made ,aside from the obvious physical differences .
Hopefully we can get back to a more productive discussion soon about improving the software.

[Ghislain]

Their marketing team still has their job to do, the product description still mentions FY6600 everywhere (except for the Frequency table) 
http://en.feeltech.net/html/Signal%20generator/show_61.html

[DaveR]

That is interesting, I wonder how it affects the operation of the device to leave the trip state buffer enable/disable pin floating. It doesn't mention that anywhere in the datasheet

Note 5 on page 1 of the datasheet I downloaded states, "Output is enabled with no connection to Pad 8. Leave Pad 8 open if disable function is not required." 

I wish I'd spotted that when I did mine!

[SMB784]

That is interesting, I wonder how it affects the operation of the device to leave the trip state buffer enable/disable pin floating. It doesn't mention that anywhere in the datasheet

Note 5 on page 1 of the datasheet I downloaded states, "Output is enabled with no connection to Pad 8. Leave Pad 8 open if disable function is not required." 

I wish I'd spotted that when I did mine!

Thanks, I totally missed that tiny little note!

[andrewjessop]

Hey everyone,

What an interesting yet very long thread! I've just read through 50pages from beginning to end and finally finished

Two questions from me:
1. Is there any difference between the 15Mhz and 60Mhz versions - from both a hardware and FPGA bitstream point of view?
2. Any updates on the custom firmware?  Seems like everyone has gone quiet.

Cheers,
AJ

[soundtec]

Hi AJ,
Physically the hardware is identical if you have a 15 or a 60 meg unit ,what changes is the firmware in the front panel ,that allows you access the higher frequencies on the FPGA board .

Yeah the thread has gone very quiet of late , as you can see from reading the entire thread ,some of the earlier units gave trouble ,theres around half a dozen still waiting on a fix ,Anything above firmware 3.2(which is what I have ) so far has worked fine as far  as I know .

As you will also see the model FY6800 is about to hit the shelves ,several of the suggestions made here have ended up incorprated into it ,like a three pin mains connection ,rubberised buttons on the front panel instead of polycarb bubble type. The firmware will also updateable by the user . No one has actually got their hands on the unit yet so probably some other surprises to come as well.

I was using the unit recently to generate audio tones and I noticed that on certain combinations of frequency and amplitude the aliasing previously described due to clocking frequency could be heard as an overtone or distortion on a relatively low audio frequency sine waves .Thats really our .5% thd limit ,  its got nothing to do with distortion in the output stages . Exactly as previously described the anomaly comes and goes depending on frequency the unit is set too ,but maybe the amplitudes of the original samples plays a role too . Maybe theres something simple we could do with the waveforms used to generate the tones to mitigate the aliasing issue .I think this is the weak point of the hobbist type sig gens vs the ones that cost many hundreds or thousands .Its a logic problem and I think the right approach in software can do a lot to help.

What would be amazing would be a auto calibration mode in software ,link each channel to the input ,run the calibrator and bang everything is needle flat from dc to 60mhz,and offsets are tweeked to avoid bad numbers from the dacs too .

Anyway hopefully the software side of this project continues ,I think all of the hardware mods are fairly well documented.

 

[cybermaus]

The firmware will also updateable by the user

What makes you think that? Surely not that Chinglish promise at the start?

It is my firm believe it is not update-able, and the fact they did not respond to several questions about it in the FY6800 giveaway thread seems to confirm that.
Pretty sure it is a near identical device electrically, and I hope identical enough for a shared firmware project.

And yes, the custom firmware is a bit quite of late, but its a one man show, and its been good weather outside in France for the last few weeks, surely fremen67 has been positively distracted. I do hope though, in case he is completely busy with other stuff, he publicizes his sources. But even then, who else is going to spend all the time...

[fremen67]

Here is a new version v0.6 of the bluepill firmware with the sweep function, as well as the corresponding modified PC Software v0.6.

I did not gave up for those wondering  . I am still working on it during my free time.

Sweeping was not a big problem but sweeping fast and accurately was not that easy.
Counting 999.99s in a 32bit word allows a minimun resolution of 10µs. That was my starting point.
After some optimization, sending a command through SPI @18Mhz  takes 5.6µs. My goal was then to be able to update a new value (Freq, Amp, Offset or Duty) in less then 10µs, that is computing the next value in less than 4.4µs where the original FY6600 updates values every 900µs...
 I had to face some more or less challenging problems, the biggests beeing:
- CPU has only 16bit counters
- with the standard C library a 64/32 bits division takes about 40µs (at least one needed)
- with the standard C library the x^y function takes more than 100 µs (one needed for logarithm  mode)
Plus the sweep parameters are not readable with the Feeltech protocol and the sweep functions in the PC Software were a pity so I had to rewrite them.
But it would have been too easy otherwise 

So, as said, the sweeping time resolution is 10µs.
In sweeping mode, a new value can be sent every 9.2µs in linear mode and every 12µs in logarithmic mode.
Sync mode works well in sweeping mode with duty (add 5.6µs) but not so much with Frequency as the phase is not very stable at high frequencies.
I did not code Amplitude and Offset Sync in sweeping mode at the moment (with calibration values beeing different between channels , I can't send the same values for both channels, plus you can have different output limits between channels). May be later on if this is usefull.
I may also add later on calibration values for VCO IN.
I also wanted to add more diagnostic information on PC Software and modify init functions but this will be for the next time...
Also everything is ready for a sweeping function on channel 2. I will just need 2 more STM32 Timers that I prefer to have in spare for now.
I will now start the modulation module.
Any volonteer for an extensive quality check on the sweep module?
Thank you for your feedback!

[DerKammi]

^^HERO^^

I so want to test this function and sweep some 10.7MHz filters I have for my AKAI AT2400 Tuner. Vero board time for a opamp output buffer then

Thanks a lot for you effort again. I've off for a while but I need to update the schematic for the counter part. Still want to make a custom PCB for the PS though.

[cybermaus]

I am very busy this week (travel for work, visiting customers. All that annoying bread-on-the-table stuff.)
So alas, I may not give fast feedback.
However, I did last week get my new LA, so when I eventually do get to it, I will try that too.

(In fact, visiting your home: France, French Alps, Annecy, probably not even time for the Mont Blanc though)

[plb]

I lost the negative 5V regulator for the second time. Now replaced with TO220 package as I did previously when I lost the positive 5V regulator. However the CH2 amplifier had also gone. It seems to be an AD8009, SOT23 package marked HJK. The replacement is on its way.
The attached photo shows the two AD8009s, each with two pots adjacent. Maybe these are for trimming the positive and negative half cycle amplitude, but I'm not sure enough to blunder in.
Does anyone know the function of these pots?
(Edit) The CH2 AD8009 on the right has been removed, pending replacement

[rhb]

I hope no one will be upset if I ask a few questions rather than search through the entire thread.  I have a V 3.1 unit I'd like to get working properly.  I now have some blue pills and STM-Link programmers and the connectors to fix the cable I borked.   I'm sure that the information is somewhere, but I've not been following the thread closely.

I got busy buying test gear which required a major reorganization of my work bench.  I still need to complete the new power distribution system as power strips became impractical because of the number of cords going to them.  Let's just say I decided to buy the bench of my dreams for my 65th birthday.  So I now have an HP 8560A 9 KHz to 3.2 GHz SA with TG among other toys.

What is the the granularity of the sample clock?  What sample rates are available for the output?

How long does it take to change the output sample rate?  That is, how quickly can the sample rate be changed?  The recent post by fremen67 suggests that we can change the sample rate in less than 1/3 the time it takes to output a full waveform.  But I may not understand what he means by updating in 10 uS.

I'd like more information about the 0.5% THD limit.  At what frequencies and sample clock rates are the spurious tones being created?

I've downloaded everything from github, but have not started reading it yet.  I'll start that this evening.

[cybermaus]

I am not fully understanding your sample questions.

But Fremens67 software cannot (yet) be loaded onto the front-board. Only on bluepill
However, if you have a 3.1 unit, then all you have is a broken sine wave, right? Not the full corrupted screen. If so, all you need to do is flash the WinBond chip. Instructions for that are around page 40 or so. Your STM32 is still good. (or at least, I do not remember anyone reporting anything other then sine wave corruption on V3.1)

You'd need to update Winbond anyway, because even if you want to go BluePill, we discovered Fremens firmware only works with V3.2 of the WinBond. Its also posted on this thread somewhere, a couple of pages after the Winbond programming instruction.

[rhb]

I'm sorry.  I've got V 3.0.   Will the blue pills not work with that?

My sample questions are related to the sweep feature.  I thought I'd see if I can contribute something to that.  Is fremen67's  stored waveform swept? I was thinking of writing a program that would  calculate waveforms so that one could load a couple of waveform registers alternately so that one got a smooth sweep by changing the sample rate and waveform register.